Hi there! I’d like to share my project with you all.
What is this? Vigil is a lightweight, self-hosted dashboard that watches your Docker images and tells you when updates are available. It’s a ready-to-run Docker setup with a simple install scripts. I know most people don’t like scripts, but since I’m a tech noob I find it pretty useful. For all the pros out there, you can check the script by yourself. This is my first “real world” project so it might not be as polished as other apps out there. It’s a hobby that I started cultivating a few months ago and I’m pretty excited with the results. However, it’d only mean something significant, if other people use it and give their own opinions about it.
If you have a few minutes, I’d really appreciate you trying it out and leaving a review or suggestions on the repo or even here. I’d do my best to answer most of the comments.
Edited because the link wasn’t showing up and giving more details about the project. https://github.com/kumucode/vigil.git
Well, I’m no tech expert at all so I’m just trying to get things right. I might not be able to answer everything, but I’ll do my best to get you an answer.
Have you vibecoded this?
Absolutelly vibecoding it with Cloude. I understand a bit of python and html but I’m no dev or technical professional at all. I just wanted to see if I could build something useful without much of technical expertise.
I appreciate you being honest in your response here.
I’d recommend adding this disclaimer to the post text and repo readme for complete transparency, and so anyone who doesn’t want to use AI-generated projects can move on without creating arguments in the comments.
There are many genuine reasons to not trust code generated by LLMs, especially with anything network-connected or handling important data, so it’s important to be upfront about it.
I think it’s awesome that you’re trying to get into larger scale software development. Agentic coding can do some amazing stuff, but it takes experience and knowledge to keep it going down good path. I think this can be a good learning opportunity to level up your own skills. Something I would suggest doing is instruct Claude with something like:
You are an experienced Senior Software Engineer that is an expert in web and backend technologies like Python, Typescript, Node and React. You are being brought in to analyze and productionize a prototype application. Please explore this project and plan out a workstream to level up this prototype so that it is production ready. First you should establish some research topics and write them to "docs/research/{date}-{topic-name}.md". After that, launch some FOREGROUND general-purpose agents to handle researching these topics in parallel. Once completed these general-purpose agents should write their findings to their original docs/research/{date}-{topic-name}.md.Once it’s conducted all the research, take a look at the documents that it writes. And if you have questions about the research results/decisions, have Claude explain.
Hi @ramielrowe thanks for the feedback, that’s actually pretty good and I’ll start using it. I understand that all this AI thing can be sloppy, and create more friction than good, but I’m really fascinated by how it can help people with little knowledge to build something that a few years ago would’ve been only possible by experts.
100%, just take a quick look at the repo. I wish there was a rule in this community that requires a label for vibecoded apps.
It’s not a bad idea at all to have a label so we could set expectations right. But don’t be too harsh on me ;) Just being able to pull a functional app without much of experience is already a reasonable accomplishment is it?
Why would this be an accomplishment we need to celebrate? Something else then you wrote that code. If you want to celebrate an accomplishment you could say “I was part of an AI vibe coding project and we created something functional”. What you did now was putting yourself front and center where you have no place to be, you are a supporting actor, at best. Its like a project manager telling everyone they accomplished getting a product out the door, giving people the idea they did that by themselves only. No, you were part of a team where (most probably) the real work was done by others. Same applies here: you used the coding abilities of another/something else to somehow toot your own horn and tell the world you did this. You did not. You never shared any info on the others involved on your team who did all the heavy lifting, only to reveal this info when pressed by others.
I get your point about giving proper credit to the tools involved, and that’s fair. I’m not trying to pass this off as traditional from-scratch coding. Reducing it to “you did nothing” feels a bit excessive. At the same time, there’s still effort in figuring out what to build, iterating, debugging, and getting something functional out. That’s the part I’m happy about.
Even this comment stinks of LLM style. Please stop trying to bring about the dead internet.